Method at a glance
- Modality: multimodal (validated in nociception and respiroception; in principle any gradable aversive interoceptive stimulus)
- Measures: The magnitude a person anticipates an upcoming interoceptive stimulus, the magnitude they then experience, and the signed difference between the two — separately on sensory, fear and affective dimensions. The Discrepancy variant additionally measures whether a person registers a violation of a learned bodily expectation.
- Procedure: Familiarization: three shades of grey are paired with three stimulus intensities, each pairing shown twice (once ascending, once randomized), with the three VAS introduced after each. Accuracy of Interoceptive Anticipation (AIA): 18 trials, each cue shown 3 s; a stimulus follows on 1/3 of trials and the participant rates what they experienced; on the other 2/3 no stimulus arrives and the participant rates what they had anticipated. Interoceptive Discrepancy (ID): 18 trials; the anticipation rating is taken immediately after the cue and before any stimulus; 2 s later a stimulus arrives that matches the cue, is one level stronger, or is one level weaker (1/3 each); the experience rating follows. Discrepancy = experience minus anticipation, so negative values are overestimations and positive values underestimations. All ratings on three VAS: sensory (0-10 intensity), fear (0-10), affective (-10 pleasant to +10 unpleasant, conventionally reverse-coded so that positive = more unpleasant).
- Also called: AIA paradigm, Accuracy of Interoceptive Anticipation, Interoceptive Discrepancy paradigm, ID paradigm, discrepancy paradigm, anticipation paradigm
Strengths
- Anticipation is measured directly, before the fact. In the parent Inspiratory Breathing Load (IBL) paradigm paradigm the anticipation rating is taken only after the moment the stimulus would have arrived, so it is contaminated by relief at its non-arrival. The ID paradigm takes it pre-stimulus. Wilzok et al. found no difference in anticipation magnitude between the two timings and still recommend the pre-stimulus version, on the correct grounds that a prospective measure is prospective whether or not it happens to agree with a retrospective one.
- Graded, not present/absent. Three cue levels let the cue→anticipation and stimulus→experience functions be traced rather than assumed, which the single-load IBL cannot do.
- Expectation violation is manipulated, not merely observed. The ±1 shift makes discrepancy an experimental effect with a known ground truth, rather than an individual difference read off spontaneous mismatch. This is what lets the paradigm index awareness of a violation.
- Three rating dimensions from every trial — sensory, fear, affective — which separate intensity from threat appraisal from hedonics. In practice sensory ratings run highest and fear lowest in both channels.
- Channel-portable by construction. The same cues, scales, trial structure and timing are used for pain and for breathing, which removes the modality/instrument confound that has dogged cross-channel comparison — the confound the HRDT/RRST pair was built to remove for the cardiac/respiratory axes.
- Reliable enough to use on individuals, tentatively: anticipation and experience ratings retest at ICC 0.66-0.92 over ~7 days, induced discrepancy at 0.78 (respiroceptive) and 0.54-0.66 (nociceptive).
Limitations
- It is a self-report difference score, not a psychophysical measure. There is no threshold, no slope, no d’, no correct answer at the trial level. A participant’s discrepancy value is the gap between two VAS marks; nothing establishes that either mark tracks the physical stimulus better than a demand-consistent response strategy would.
- Difference scores inherit the noise of both terms and are the classic vehicle for shared method variance. Because both ratings use the same scales in the same session, a cross-channel correlation of discrepancy is explicable by consistent scale use alone — the standing objection to the r = 0.57 in the validation paper.
- The cue is exteroceptive and arbitrary. What is learned is that a grey patch predicts an experimenter action. This is ordinary associative learning whose outcome happens to be interoceptive, not a model of the body’s own dynamics — see Forecasting (internal-state prediction), where the same distinction disqualifies the BLT.
- Three levels will not identify the psychophysical function. Linearity is assumed over the tested range and untested outside it.
- Intensity matching across channels is not solved. In the validation study the respiroceptive stimuli spanned the full VAS while the nociceptive ones clustered low, so the two channels were not equated on the dimension that matters for comparing them.
- Stimuli are hand-applied by an unblinded experimenter (PinPricks, resistors swapped on tubing), so the design does not exclude cue-consistent application drift.
- Aversive by design, which bounds the populations it can be used in and means every measurement is taken under perturbation — a virtue for one question and a confound for others.
Connections
Two paradigms introduced together by Wilzok, Adamic, Khalsa & Croy (2023), described on one page because they share a familiarization phase, a stimulus set, a rating instrument and a purpose, and differ in exactly two design decisions.
The two decisions that separate them
| Accuracy of Interoceptive Anticipation | Interoceptive Discrepancy | |
|---|---|---|
| when anticipation is rated | retrospectively, on the 2/3 of trials where no stimulus arrived | prospectively, after the cue and before every stimulus |
| what stimulus follows the cue | the cued intensity, on 1/3 of trials | the cued intensity, or ±1 level, at 1/3 each — on every trial |
| what it yields | anticipation and experience as separate quantities, never on the same trial | a within-trial anticipation→experience pair, and an experimentally induced mismatch |
| what it cannot do | speak to expectation violation at all | give an uncontaminated anticipation estimate for an un-violated expectation after the first few trials, since two-thirds of trials teach the cue is unreliable |
That last cell is the paradigms’ unresolved tension and neither the authors nor this wiki can currently settle it. The ID paradigm’s manipulation works by making the cue misleading two-thirds of the time. A participant who notices — and the whole paradigm rests on their noticing, since that is what discrepancy awareness is — should rationally regress their anticipations toward the middle as the block proceeds. Trial-order effects on anticipation ratings are not reported.
What they are for
They exist to give Paulus & Stein’s clinical thesis an instrument. If psychopathology is a failure to appropriately anticipate interoceptive states, then anticipation has to be measurable as a graded quantity, separately from the experience it is an anticipation of, and the mismatch between them has to be quantifiable. That is precisely the shape of these tasks, and it is why the Interoceptive psychopathology programme is the stated target application (anxiety, depression, anorexia, fibromyalgia).
The measured quantity is not precision in the psychophysical sense, and should not be reported as such. It is a signed magnitude gap between two reports.
Placement among the wiki’s methods
The wiki’s interoceptive tasks divide by what they hold fixed:
- Threshold instruments (FDT, RRST, HRDT) fix the judgement — was there a signal — and vary the stimulus to find where perception fails.
- Attention paradigms (Interoceptive attention task) fix nothing and have no correct answer; they measure where attention goes.
- Learning paradigms (BLT) vary a cue-outcome contingency and fit a learning model to the trial sequence.
- These paradigms fix the physical stimulus set and vary the expectation, then read the gap. They are the only ones in the wiki whose dependent variable is a mismatch between a forecast and an outcome.
The nearest neighbour is the BLT, and the difference is instructive: Harrison et al. fit a formal learning model to trial-by-trial predictions and recover prediction certainty and prediction error as model parameters; Wilzok et al. take the participant’s own numerical report of the gap. The BLT is computationally richer; these paradigms are cheaper, run in two channels, and ask the person rather than the model.
The nociception question
Half of the validated application is pinprick pain, which raises the Nociception scope problem directly: if only visceral nociception counts as interoceptive (Ceunen, Vlaeyen & Van Diest 2016), then one of the two channels here is somatosensory and the paradigm’s cross-channel claim is not an interoceptive one. The authors take the broad Craig-style line explicitly. Anyone using these paradigms inherits that commitment and should state it.
What would establish them
- A clinical sample. The paradigms were built for Interoceptive psychopathology and have never been run in it. The whole rationale is untested.
- Relation to self-report. MAIA-2 was collected in the validation study and never reported against task measures. Whether anticipation accuracy relates to sensibility is the obvious first question and remains open.
- A third channel, and a resting comparison. Adding cardiac or gastric would test whether the r = 0.57 survives; running a non-aversive variant would test whether it is perturbation that produces it (see Is interoception domain-general or organ-specific?).
- Trial-order analysis, for the reason in the table above.